Rainfall Variability Assessment—A Case Study of Rokel-Seli River Basin in Sierra Leone

Abstract

Rainfall is a most important factor for climate study. It is depending on various factors of hydrological cycle. Actual projection of rainfall pattern is quite difficult due to involvement of uncertainty with respect to space and time. Hence, it is very difficult to access actual occurrence of rainfall in daily basis. For climate study various agencies are developed number of climate data base as per long term historical hydro-meteorological data sets. Using these climatic data base, one can project the rainfall patters under particular uncertainty bands and understand the rainfall variability at particular space. Understanding the trend and variability of rainfall is also important to determine the supplemental water requirements for crops as well as water resources planning during their critical water deficit periods. The aim of this study is to investigate the past variation of rainfall and to identify the trend over Rokel-Seli river basin in Sierra Leone for a period of 45 years (1961–2005) of rainfall data. Rokel-Seli river basin is importance to the country’s economy as it supplies water to the Bumbuna hydroelectric power scheme as well as water for the agriculture, fisheries, mining and transportation and for ecological purposes. The long-term trend has been detected using the Mann Kendell (MK) and Modified Mann Kendall (MKK) test(s) for historical time series in terms of monthly, seasonal and annual basis. Further, shift change point has been detected for break point identification using SNHT and MWP test(s). Moreover, rainfall has been projected till 2050s under different climate scenarios with various CMIP5 emission conditions, i.e., RCP-2.6, RCP-4.5 and RCP-8.5. Project rainfall and its trends can be useful for future prospective of agriculture and water resources planning and mitigations under consideration of Climate change.

25.1 Introduction

Rainfall variability assessment is quite important for determining the water availability and requirement for a catchment which can influence water use and allocation. Precipitation is most important factor which influences the living beings directly and indirectly. However, precipitation play key role in hydrological cycle (Prabhakar et al. 2019). The overall hydrology is initiated with precipitation and is regulate hydrology, vegetation and water bodies, and it is also significant for agricultural production (King et al. 2014). The occurrence and variability of precipitation influences to a large extent which crops can grow in different areas/regions throughout the world (Silberstein et al. 2012). A good knowledge of rainfall variability is relevant to help optimize farm production in a sustainable manner (Gajbhiye et al. 2015).

The Rokel-Seli river basin is of critical importance to the country’s economy as it supplies water to the Bumbuna hydroelectric power scheme as well as water for agriculture, fisheries, mining and transportation and for ecological purposes. As per future prospective, quantum of water availability from precipitation is also quite important (Zhao et al. 2008). Hence, allocation of water for irrigation, industrial and domestic use is also very important and interlinked with hydrological cycle. Understanding the trend and variability of precipitation is also necessary to determine the supplemental water requirements of crops during their critical growth periods (Liu and Lin 2004). The response of hydrologic circulation to climate and land use changes is important in studying the historical, present, and future evolution of aquatic ecosystems.

The present study highlights the rainfall variability over the Rokel-Seli River Basin in Sierra Leone. The adopted analysis provides key information of basin’s water availability and hence this work offers benchmark information that can be used to increase the capacity of long-range water resource planning and management, land use planning, agricultural water development and conservation, and industrial water use over the next several decades at basin level. Climate change impacts the rainfall distribution all around the world. The variation in rainfall distribution would alter the storage, recharge surface runoff and soil moisture (Masafu et al. 2016). Moreover, the rainfall variation can increase or decrease the discharge and water availability to a river basin. Increased variation in the intensity and frequency of precipitation is one of the major impacts of climate change (Anandhi et al. 2008).

25.2 Description of Study Area

The Rokel-Seli River, which is the largest river in Sierra Leone is 356 km (221 miles) in length and has a width varying from 6.4 to 16.1 km (4–10 miles). It stretches across the entire northern region before joining the Atlantic Ocean. It has a basin area of 10,699 km² which infringes four major districts (Koinadugu, Bombali, Tonkolili and Port Loko districts) having 31% (2,159,119) of the total country population (7,075,641) of Sierra Leone. Location map of study area is shown in Fig. 25.1.

Fig. 25.1

Location map of study area

This basin is characterized by a heterogeneous forest-savanna mosaic and experiences a humid tropical climate with annual rainfall averaging 2435 mm and mean monthly temperature of 27.78 °C. Elevation difference is about 956 m and varies from 19 to 975 m. There are two main seasons: rainy season (May–October) and dry season (November–April). There are several small traditional villages in the area with rice cultivation in wet depressions and harvesting of non-timber forest products such as oil palm nuts. Between the upstream and downstream, lots of mining and agricultural activities are taking place such that the rainfall pattern within this area varies considerably.

25.3 Methodology

25.3.1 Assessment of Rainfall Variability

In other to assess the rainfall variability, the rainfall data used for the thirteen (13) stations in the Rokel-Seli river basin from 1961 to 2005 (45 years) was obtained from the Sierra Leone water security website (https://www.salonewatersecurity.com/data) which was created to serve as a repository for hydrological (rainfall, surface water and groundwater) data. This was achieved by the Ministry of Water Resources, the lead government institution of Sierra Leone responsible for monitoring water resources in collaboration with multiple and diverse organizations re-established for hydrological monitoring activities. The extracted monthly, annual and seasonal rainfall data for the entire basin (station-wise) can be shown in Table 25.1.

Table 25.1 Historical average monthly, annual and seasonal rainfall (mm) of RSRB for the period of 1961–2005

The average annual rainfall for the basin was estimated as weighted rainfall. The area covered by each rain gauge station with their corresponding annual and seasonal rainfall values were computed to give the mean annual rainfall of RSRB as shown in Table 25.2.

Table 25.2 Mean annual and seasonal weighted station rainfall (mm) of RSRB for the period of 1961–2005

The mean annual and seasonal depth of rainfall of Rokel-Seli river basin was estimated to be 2435, and 2190 and 245 mm in the wet season and dry Season respectively as illustrated in Fig. 25.2.

Fig. 25.2

Average annual and seasonal rainfall distribution over RSRB during 1961–2005

This means that 90% of the average annual rainfall occurs in the wet season with only 10% contribution of rainfall in the dry season on the basin. The assessment of rainfall variability is of crucial importance for stakeholders and policy makers to provide information for an improved water management. Table 25.3 gives the statistics relating to the variability of the annual and seasonal rainfall of Rokel-Seli river basin.

Table 25.3 Average annual and seasonal rainfall Statistics of RSRB for the period of 1961–2005

Considerable aerial variation exists in the annual rainfall on the basin with highest rainfall of magnitude 3589 mm annually and, 3336 and 715 in the wet and dry season respectively.

Based on the rainfall variability analysis carried out on the historical rainfall, it has been detected that the average annual rainfall over the basin was 2435 mm, varying from minimum rainfall of 895 mm to 3589 maximum rainfall. Most of the rainfall occurs during the wet season, contributing 90% of the annual rainfall over the basin ranging from minimum rainfall of 844 mm to 3336 mm maximum rainfall with average annual rainfall of 2190 mm which occurs during the months of May–October. The dry season which occurs between the periods of November to April has only a maximum rainfall of 715 mm contribute only 10% of annual rainfall on the basin with a minimum of 11 mm with mean annual rainfall of 245 mm (Table 25.3).

The coefficient of variation of the annual and seasonal rainfall varies between 14 and 52 with an average value of about 25. The coefficient of variation is least at stations of high rainfall and largest in regions of scanty rainfall as indicated in Fig. 25.3 at station Seng be (upstream) and station Yoni (downstream) for both annually and seasonally.

Fig. 25.3

Inter-annual variability of rainfall (%CV) for a annual, b wet and c dry season over RSRB during the period of 1961–2005

This analysis can be useful to detect the changes in the precipitation-streamflow relationship and quantifies the impact of precipitation to runoff (i.e. response of streamflow to climate change) in the basin. It was observed that two stations; one at the extreme upstream and the other at downstream gives highest values of coefficient of variation ranging between 30 and 52% for both annual and seasonal. This means, less rainfall occurs at those points on the basin during the period of 1961–2005. There is less variation in most part of the basin considering the middle part ranging between 13 and 16% noting that, much of the rainfall is being received at this point. The absolute variability of rainfall distribution was 256 and 232 annually and wet season for the entire basin (Table 25.3); noting that there is significant magnitude of deviation from their mean values of rainfall. However, the standard deviation for the dry season was found to be 45 depicting that, there is low dispersion or variability from the mean rainfall. Hence, there is little or no significant amount of rainfall in the dry season.

25.3.2 Rainfall Trend Analysis

Trend analyses of precipitation are useful to explore the impact of climate change on water resources. Thus, trend analyses of precipitation are required to facilitate more sustainable water planning and management. In this study, the long-term trend has been detected using the Mann–Kendall (MK) test for historical time series in terms of monthly, seasonally and annually.

25.3.2.1 Mann–Kendall Test (MKT)

The non-parametric Mann–Kendall test is commonly employed to detect monotonic trends in series of environmental data, climate data or hydrological data (Hamed and Rao 1998). The null hypothesis, H₀, is that the data come from a population with independent realizations and are identically distributed. The alternative hypothesis, H_A, is that the data follow a monotonic trend (Pohlert 2016). In this study the Mann–Kendall test (MKT) and the Modified Mann–Kendall test (MMKT) has been used to detect the rainfall trend during the period of 1961–2005 over RSRB. The Mann–Kendall test statistic S is calculated using the formula as follows:

$$ S = \sum\limits_{i = 1}^{N = 1} \cdot \sum\limits_{j = i + 1}^{N = 1} {{\text{sgn}} \left( {x_{j} - x_{i} } \right)} $$

(25.1)

where $x_{j}$ and $x{}_{i}$ are the annual values in years j and i, j > i respectively, and N is the number of data points. The value of ${\text{sgn}} \,(x_{j} - x_{i} )$ is computed as follows:

$$ {\text{sgn}} \left( {x_{j} - x_{i} } \right) = \left\{ {\begin{array}{*{20}c} 1 & {if} & {\left( {x_{j} - x_{i} > 0} \right)} \\ 0 & {if} & {\left( {x_{j} - x_{i} } \right) = 0} \\ { - 1} & {if} & {\left( {x_{j} - x_{i} } \right) < 0} \\ \end{array} } \right. $$

(25.2)

These statistics represents the number of positive differences minus the number of negative differences for all the differences considered. For large samples (N > 10), the test is conducted using a normal approximation (Z statistics) with the mean and the variance as follows:

$$ E\left[ S \right] = 0 $$

(25.3)

$$ Var\left( S \right) = \frac{1}{18}\left[ \begin{gathered} N\left( {N - 1} \right)\left( {2N + 5} \right) - \sum\nolimits_{p = 1}^{q} {\; \cdot } \;t_{p} \left( {t_{p} - 1} \right) \hfill \\ \left( {2t_{p} + 5} \right) \hfill \\ \end{gathered} \right] $$

(25.4)

Here q is the number of tied (zero difference between compared values) groups, and $t_{p}$ is the number of data.

values in the pth group. The values of S and VAR(S) are used to compute the test statistic Z as:

$$ Z = \left\{ {\begin{array}{*{20}c} {\frac{S - 1}{{\sqrt {Var\left( S \right)} }}} & {if} & {S > 0} \\ 0 & {if} & {S = 0} \\ {\frac{S + 1}{{\sqrt {Var\left( S \right)} }}} & {if} & {S\left( S \right) < 0} \\ \end{array} } \right. $$

(25.5)

The presence of a statistically significant trend is evaluated using the Z value. A positive value of Z indicates an upward trend and its negative value a downward trend.

25.3.2.2 Modify Mann–Kendall Test (MMKT)

The Modified Mann–Kendall test has been used for trend detection of autocorrelation series. Therefore, in this analysis the autocorrelation between the ranks of the observations ‘pk’ has been estimated after subtracting the non-parametric Sen’s median slope from the slope.

$$ \begin{aligned} & \frac{n}{{n^{ * } }} = 1 + \frac{2}{{n\left( {n - 1} \right)\left( {n - 2} \right)}} \times \sum\limits_{p}^{q} {\left( {n - k} \right)\left( {n - k - 1} \right)} \\ & \left( {n - k - 2} \right)p \\ \end{aligned} $$

(25.6)

Significant values of ‘pk’ have only been used for calculating the variance correction factor n/n* and it was calculated from the equation proposed by Hamed and Rao (1998).

where:

n represents the actual number of observations,

n* is represented as effective number of observations to account for the autocorrelation in the data and pk is the autocorrelation function for the ranks of the observations.

The corrected variance is then given as (Hamed and Rao 1998).

$$ Var^{ * } \left( S \right) = Var\left( S \right) \times \frac{n}{{n^{ * } }} $$

(25.7)

where $Var\left( S \right)$ is from Eq. (25.4).

25.3.3 Magnitude of Rainfall Trend (Theil Sen’s Slope)

Sen (1968) developed a non-parametric method to estimate the magnitude (slope) of the trend in a time series (Sen 1968). This method assumes a linear trend in the time series. In this method, the slope Qi of all data value pairs are calculated according to:

$$ Qi = \frac{{x_{j} - x_{k} }}{j - k} $$

(25.8)

where j > k. If there are n values $x_{j}$ in the time series we get as many as $N = \frac{{n\left( {n - 1} \right)}}{2}$ slope estimates Qi. The Sen’s estimator of the slope is the median of these N values of Qi. The N values of Qi are ranked from the smallest to the largest and the Sen’s estimator as follows:

$$ Q = \left( {Q_{{\left[ {\frac{n + 1}{2}} \right]}} } \right),\quad {\text{if N is odd}} $$

(25.9)

Or

$$ Q = \frac{1}{2}\left( {Q_{\frac{n}{2}} + Q_{{\left[ {\frac{n + 2}{2}} \right]}} } \right),\quad {\text{ if N is even}} $$

(25.10)

The two-sided test is carried out at 100 (1 − a) % of the confidence interval to obtain the true slope for the non-parametric test in the series. The positive or negative slope Q_i is obtained as upward (increasing) or downward (decreasing) trend. In the present study, the test was carried out at 5% significance level, therefore when Z value exceeds ±1.96 null hypotheses is rejected and show the existence of trend in the series as in Table 25.4.

Table 25.4 The sen slope and Z-statistic values of annual and seasonal rainfall using MK and MMK for RSRB during 1961–2005

The Z-statistics value was analyzed (Chandniha et al. 2017) as follows:

• −1.96 < Z < 1.96 = No Trend (Not significant)

• Z > 1.96 = Increase in trend (i.e. positively significant)

• Z < −1.96 = Decrease in trend (i.e. negatively significant).

The values of +Z and −Z indicates upward and downward trend respectively. The Z values of Mann–Kendall test accept the null hypotheses of no trend when;

$$ \pm Z \le {{Z_{1} - x} \mathord{\left/ {\vphantom {{Z_{1} - x} 2}} \right. \kern-\nulldelimiterspace} 2} $$

where x is the level of significance at two tailed trend tests.

The Theil Sen’s slope estimator is a robust method of robustly fitting a line of sample points in the plane by choosing the medians of the slope of all line through pairs of the points. It has been viewed as the most popular nonparametric analysis for determining a linear trend and therefore this can be illustrated by box plot in Fig. 25.4.

Fig. 25.4

a Box plot of the Theil-Sen slopes for annual and seasonal rainfall time series of RSRB. b Box plot of the Theil-Sen slopes for monthly rainfall time series

25.3.4 Homogeneity Test in the Times Series (1961–2005)

Application of homogenization on climatic time series preserve the climatic signal and reduce the impact of non-climatic factors in the time series. Therefore, it is important to address these factors in order to develop homogenized records for studying climate change. Change-point analysis examines climate data discontinuities and it directly addresses the question of where the change in the mean value of the observations is likely to have occurred.

Homogeneity in trends was tested using the Standard Normal Homogeneity Test (SNHT) to obtain the homogeneous and heterogeneous trend in the data time series with significance level of 5%. Change point (P values) has been computed using 10,000 Monte Carlo simulations base on Mann–Whitney-Pettit (MWP) Test and SNHT (Alexanderson 1986; Alexandersson and Moberg 1997).

The test interpretation, H₀: stands for the homogeneous series and Ha: there is a date at which there is change in the data as shown in Table 25.5. As the computed p-value is greater than the significance level alpha = 0.05, one cannot reject the null hypothesis H₀.

Table 25.5 The MWP test and SNHT test for RSRB (1961–2005)

Trend analysis was done applying MK and MMK tests and Sen’s slope on monthly, annual and seasonal rainfall data for all stations in the basin. The MK and MMK statistics at 5% significance level are shown in Table 25.4. Among all the 13 stations in the basin, 3 stations show a negatively significant trend in wet and dry season and just 2 stations show positively significant trend in the dry season. From the Z-statistics value, there is not any significant annual trend in all stations, although 9 stations show negative trend and 4 shows positive trend annually during the period of 1961–2005.

Monthly and annual trends are shown in Fig. 25.4 using the box plot of Theil-Sen’s slope method for the Rokel-Seli River basin. The box central line represents median, upper and lower lines represent the 75th and 25th percentile, respectively. Also, the upper and lower lines represent the maximum and minimum values of rainfall slopes. Based on the analysis, the dry season shows a positive slope and similarly all the months found in the dry season (November–April) also shows positive slope in the monthly box plot. This denotes that, during periods of high rainfall the Theil-Sen’s slope will show negative trend and vis-à-vis during low rainfall. Using MWP and SNHT tests in identifying homogeneity in the data series (Prabhakar et al. 2018), results in Table 25.5 shows that there was no shift change point detected in the data series for the period of 1961 and 2005. Hence all P-values shows Ho (Null Hypothesis) trend (i.e. all data in the series are homogenous during the time series) (Yusof and Kane 2013).

25.3.5 Rainfall Projections

Rainfall projections are necessary in determining the water balance and future water allocation in the basin. Thus, evaluating the future variation of hydrologic cycle and water resources has special significance for regional planning and water resources management. It helps in the assessment of the future impact of climate change over the basin which affects changes in the hydrologic cycle of the basin. Global Climate Models (GCMs) are the fundamental tools that provide future projections of climate variables in the changing ecosystem. Therefore, studies dealing with the climate change impact assessment at catchment scale require downscaling of GCM projections to an appropriate scale to represent the catchment heterogeneity (Silberstein et al. 2012). Various statistical and dynamic downscaling methods have been adopted in the past to downscale large scale atmospheric variables from the GCMs to a regional scale or to a finer scale representative of a catchment (Silberstein et al. 2012; Anandhi et al. 2008). For the purpose of this study, statistical downscaling method has been applied.

25.3.5.1 Statistical Downscaling Method (SDSM)

Statistical downscaling method was applied for downscaling the monthly rainfall of RSRB for future rainfall projections using traditional downscaling regression-based approach (Chandniha and Kansal 2016); Multiple-Linear Regression (MLR). The general formula of MLR is written as:

$$ Y^{MLR} = \alpha + \sum\limits_{i = 1}^{n} {\beta_{i} X_{i} } + \varepsilon $$

(25.11)

where, Y^MLR is the estimated predictand (rainfall), α is the intercept; β is the regression coefficients, X_i is the predictor (26 parameters and ε is the error term. Regression coefficients at 95% confidence level were estimated with Durbin-Watson technique for residuals estimation. SPSS (Ver. 24) was applied for model fit, correlation values, coefficient of determination, descriptive statistics etc.

The following steps were used to carry out the methodology for rainfall projections:

• 1. Perform consistency check of observed monthly rainfall (Predictand) using Hydrognomon (Ver. 4) for the period of 1961–2005

• 2. Transfer National Centre of Environmental Predictions (NCEP)and Global Circulation Models (GCM) predictors for the study area from Canadian Centre for Climate Modeling and Analysis (CanESM2) https://www.climate-scenarios.canada.cac corresponding to Representative Concentration Pathways (RCP2.6, RCP4.5 and RCP8.5) emission scenarios and then convert the predictors from daily to monthly basis taking average of each predictor over the month.

• 3. Identify calibration period (from 1961 to 1990) and validation period (from 1991 to 2005) for the calibration and validation of the model

• 4. Develop the empirical relationship between historical rainfall (predictand) and the 26 predictors using MLR technique on calibration period data

• 5. Using expert opinion based on scatter plot, partial correlation, correlation etc. the most suitable predictors were identified

• 6. Rainfall (predictand) were estimated during the validation period using the selected predictors and compared with the observed values

• 7. The probable error in the observed and the estimated values were calculated and bias correction was applied to correct the predicated values

• 8. The important indicators of the goodness of regression were checked by the following parameter; Nash-Sutcliffe Error Estimate (NS-EE), Coefficient of Correlation (CC), Normalized Mean Square Error (NMSE), and the Root Mean Square Error (RMSE)

• 9. The suggested series of NCEP Corresponding to RCP2.6, RCP4.5 and RCP8.5 scenarios and the selected predictors were used to generate the future series of the predictand for the periods of 2020s and 2050s.

In applying the above methodology for the future projection of rainfall over Rokel-Seli river Basin the daily observed predictor data of atmospheric variables derived from NCEP 2.80° (latitude) × 2.80° (longitude) grid-scale for 45 years (1961–2005) were obtained from CanESM2. The data was extracted between latitude 8.90° N to 9.30° N and longitude −11.98° W to −11.5° W (BOX_125X_36Y). The Representative Concentration Pathways (RCP2.6, 4.5 8.5) emission scenarios were also downloaded from CanESM2. Full descriptions of NCEP variables (predictors) are elaborated in the Table 25.6.

Table 25.6 Variables and description of NCEP and GCM predictors (26 variables)

The Geostrophic air flow velocity, Vorticity, Zonal velocity component, Meridional velocity, Divergence and Wind direction are variables derived using the geostrophic approximation at different atmospheric levels.

The vorticity measures the rotation of the air, Zonal velocity component is the velocity component along a line of latitude (i.e. east–west), Meridional velocity component is the velocity component along a line of longitude (i.e. north–south), Divergence relates to the stretching and outflow of air from the base of an anti-cyclone. Wind direction variable is the only variable which is not normalized. The same parameters are considered in comparing the results based on MLR method.

The MLR equations derived for each rainfall station can be given as follows:

Bombali: 207.1 + 24.3X₁₅ − 9.8X₁₉ + 30.3X₂₀ + 169.3X₂₂

Diang: 253.6 + 45X₇ + 68.4X₁₁ + 103.3X₂₂ + 39.5X₂₃ + 45.4X₂₄ − 45.4X₂₅

Kafe Simera: 236.5 + 29.4X₁₅ − 0.5X₁₉ + 18.4X₂₀ + 215.7X₂₂

Kholifa Rowala: 226.4 + 29.4X₇ + 41.8X₁₁ + 69.1X₂₂ + 41.5X₂₃ + 30.2X₂₅

Koya: 182.1 + 13.8X₁₅ − 0.3X₁₉ + 18.7X₂₀ + 164.2X₂₂

Malal Mara: 210.5 + 42.8X₁₅ + 12.9X₁₉ + 35.3X₂₀ + 183.1X₂₂

Masimera: 213.2 + 22.8X₇ + 40X₁₁ + 64.7X₂₂ + 39.9X₂₃ + 32.8X₂₅

Mongo: 162.1 + 25X₁₅ − 1.5X₁₉ + 24.4X₂₀ + 142.3X₂₂

SafrokoLimba: 244.6 + 68.8X₁₁ + 56.8X₂₂ + 52.3X₂₃ + 25.1X₂₄ + 29X₂₅

Sambaia: 223.8 + 25.6X₇ + 22.1X₁₁ + 184.9X₂₂ − 0.9X₂₃ − 13.7X₂₅

Sengbe: 162.2 + 25.1X₁₅ + 2.9X₁₉ + 36.8X₂₀ + 137.9X₂₂

Warra Bafodia: 227.1 + 16.5X₇ + 25.6X₁₁ + 100.6X₂₂ + 26.9X₂₃ + 37X₂₅

Yoni: 199.4 + 56.4X₁₅ − 13.6X₁₉ + 41.8X₂₀ + 158.3X₂₂

Where, X₇ = Divergence; X₁₁ = vorticity; X₁₅ = Zonal velocity component; X₁₉ = Divergence; X₂₀ = 850 hPa geopotential height; X₂₂ = near surface relative humidity; X₂₃ = Specific humidity at 500 hPa height; X₂₄ = specific humidity at 850 hPa height; X₂₅ = near surface specific humidity.

25.3.6 Calibration and Validation

The daily observed predictor data obtained from NCEP reanalysis, normalized over the period of 1961–1990 and hence this period was selected for calibration, while the validation period was selected from 1991 to 2005 to normalize the calibration and validation model. Observed and estimated rainfall during calibration and validation period is shown in Fig. 25.5 and scatter plot of both cases are represented in Fig. 25.6. The important parameters or indicators of the goodness of the regression for each of the stations during the calibration and validation periods are also shown in Table 25.7.

Fig. 25.5

Observed and estimated monthly rainfall during calibration and validation periods

Fig. 25.6

Scattered plot between observed and estimated rainfall during a calibration and b validation

Table 25.7 Important indicators of accuracy of results during calibration and validation for rainfall time series at the various stations of Rokel-Seli River Basin

During the calibration and validation periods the average monthly values were calculated for both observed and the estimated rainfall as in Table 25.8.

Table 25.8 Calibration and validation values of monthly average rainfall of Rokel-Seli River Basin

The equations derived after SDSM base on MLR approach during calibration and validation periods are applied for future projections under Couple Model Inter-comparison phase-5(CMIP5) emission scenarios; RCP 2.6 RCP 4.5 and RCP 8.5 which are used in this study.

For the purpose of this work, the projected rainfall has been categorized in various time step as; past (1961–2005), 2020s (2011–2040) and 2050s (2041–2070) scenarios as tabulated in Table 25.9.

Table 25.9 Past and future rainfall statistics under RCP2.6, RCP4.5 and RCP8.5 of RSRB

Further, climate change scenarios also help the future planning of various activities which are associated with water and dependent with rainfall over the catchment.

The forecasted rainfall is expected to help the policy makers and the stakeholders for making effective water resources planning.

It can also be observed that, the MLR model fits between the observed and estimated monthly average rainfall during the calibration and validation periods with a coefficient of determination of 0.848 and 0.905 respectively as indicated in Table 25.8. Applying the recommended MLR model, average annual rainfall estimated for the period of 1961–1990 was 2474 mm as compared to the observed rainfall of 2477 mm. Similarly, the estimated average rainfall during 1990–2005 was 2444 mm compared to 2363 mm as observed rainfall. The projected mean annual rainfall for the periods of 2020s and 2050s according to CMIP5 emission scenarios are; 2744 and 2545 mm (RCP 2.6), 2508 and 2645 mm (RCP 4.5) and 2630 and 2623 mm (RCP8.5) as against observed rainfall of 2435 mm. The overall forecasted rainfall (taking the average of all 3 scenarios) for 2020s is 2627 mm and for 2050s is 2604 mm as compared to the observed value of 2435 mm. As projected for the average rainfall over Rokel-Seli River Basin, the expected rainfall to occur in 2020s and 2050s is about 7–8% higher than the observed rainfall during the periods of 1961–2005.

25.3.7 Dependable Annual Rainfall of Rokel-Seli River Basin

It is very important to know the annual rainfall dependability in planning water resources over the basin in order to obtain the relationship between the magnitude of the event and its probability of exceedance. Hence the 75% or 95% dependable annual rainfall are of utmost concern to identify the minimum water availability.

Therefore, in this study, the dependable annual rainfall has been estimated for the past (1961–2005) and future time series of 2020s (2011–2041) and 2050s (2041–2070) as shown in Table 25.10 and Fig. 25.7.

Table 25.10 Dependable annual rainfall of RSRB for past (1961–2005), 2020s (2011–2040) and 2050s (2041–2070)

Fig. 25.7

Rainfall dependable curves for past (1961–2005), 2020s (2011–2040) and 2050s (2041–2070)

The 95% dependable rainfall for the past, 2020s and 2050s were estimated as 2034 mm, 2320 mm and 2462 mm respectively. This justifies the values of minimum average rainfall that occurred in the basin and which could be expected in 2020s and 2050s. Hence these values are very useful for future water resources development and management on the basin.

25.4 Conclusions

The rainfall over Rokel-Seli river basin varies considerably from the upstream and downstream showing high value of coefficient of variation annually and seasonally. Thus, the high values of CV at these stations shows that less rainfall occurs at these parts of the basin and high rainfall occurring most part at mid of the basin with average rainfall of 2435 mm. The non-parametric (MK and MMK) tests were used to detect rainfall trends over the basin during the period of (1961–2005). The accuracy of MMK test in terms of significance level was more precise than MK test at the same level of significance, showing only 3 stations having negatively significant trend and two stations with positively significant trends in the wet and dry seasons. No significant trends were identified annually.

The Sen’s slope magnitude varies between −17.354 and 6.951 mm/year annually in the basin (1961–2005), therefore the seasonal slopes were mostly negative for the wet season and positive for the dry season. The Mann–Whitney–Pettitt and SNHT tests were used to identify possible break points in precipitation during the 45 year period. However, from the results of the study, it can be concluded that there was no shift change point in the data series. Hence the rainfall data was homogenous for all throughout the time series for the period of 1961–2005.

Rainfall projections has been carried out in the basin for the periods of 2020s and 2050s. The average annual rainfall projected in the 2020s (2011–2040) is 2627 mm and for 2050s (2041–2017) is 2604 mm as compared to the observed value of 2435 mm. This means that there would be increase in rainfall of about 7–8% unto 2050s due to climate change. The observed dependable annual rainfall at 95% is 2034 mm and the projected dependable rainfall at 95% dependability in 2020s and 2050s is 2320 mm and 2462 mm respectively. It is very important to know the dependable rainfall in planning water resources over the basin to identify the minimum water available for various water requirement, also to determine the projected minimum rainfall which could be used for future water resources management.

The present study highlights the rainfall variability over the Rokel-Seli River Basin in Sierra Leone. The adopted analysis provides key information of basin’s water availability and hence this work offers benchmark information that can be used to increase the capacity of long-range water resource planning and management, land use planning, agricultural water development and conservation, and industrial water use over the next several decades at basin level. The results of the study also help in the assessment of the future impact of climate change over the basin which affects changes in the hydrologic cycle of the basin. Also, this study can be used as a guide to simulate rainfall variability over other basins in Sierra Leone to determine the water balance within those basins. Project rainfall and its trends can be useful for future prospective of agriculture and water resources planning and mitigations under consideration of Climate change.